Advances in System Identification and Control

The field of system identification and control is witnessing significant advancements, driven by the adoption of active learning strategies, data-driven approaches, and innovative applications of existing techniques. Researchers are exploring new methods to improve the efficiency and accuracy of system identification, such as online design of experiments and coreset selection, which enable more effective data collection and analysis. The integration of data-driven modeling with traditional model-based control is also gaining traction, as seen in the development of prescribed-time control frameworks that combine the strengths of both approaches. Furthermore, the use of machine learning and deep learning techniques, such as generative latent diffusion and parametric neural amp modeling, is opening up new avenues for efficient spatiotemporal data reduction and virtual amp modeling. Notable papers in this area include the work on Thompson Sampling-Based Learning and Control for Unknown Dynamic Systems, which proposes a novel approach to active learning-based controller design, and the paper on A Data-Driven Prescribed-Time Control Framework via Koopman Operator and Adaptive Backstepping, which presents a synergy of data-driven modeling and model-based control.

Sources

Online design of experiments by active learning for nonlinear system identification

Koopman operator-based discussion on partial observation in stochastic systems

Thompson Sampling-Based Learning and Control for Unknown Dynamic Systems

Online Coreset Selection for Learning Dynamic Systems

An approximation theory for Markov chain compression

Communication via Sensing

Data-driven Implementations of Various Generalizations of Balanced Truncation

Predictor-Based Compensators for Networked Control Systems with Stochastic Delays and Sampling Intervals

On sample-based functional observability of linear systems

Active Estimation of Multiplicative Faults in Dynamical Systems

The Rate-Distortion Function for Sampled Cyclostationary Gaussian Processes with Memory and with Bounded Processing Delay: Extended Version with Proofs

Auto-optimization of Energy Generation for Wave Energy Converters with Active Learning

Parametric Neural Amp Modeling with Active Learning

Generative Latent Diffusion for Efficient Spatiotemporal Data Reduction

A Data-Driven Prescribed-Time Control Framework via Koopman Operator and Adaptive Backstepping

Built with on top of